[ase-users] Genetic algorithm- new structures

Steen Lysgaard stly at dtu.dk
Mon Dec 14 10:44:10 CET 2020


Hi Mina,

Can you check that the structures that start re-relaxing really has a row in the db with the same gaid and the key-value-pair relaxed=1?

If that is the case you can send the main and relax scripts for the GA, otherwise it's hard to tell what is going on. Looking at your optimization log it seems that the starting structures or optimization procedure are way off.

Best regards,
Steen
________________________________________
From: ase-users-bounces at listserv.fysik.dtu.dk <ase-users-bounces at listserv.fysik.dtu.dk> on behalf of Mina Taleblou via ase-users <ase-users at listserv.fysik.dtu.dk>
Sent: Monday, December 14, 2020 9:26:33 AM
To: ase-users at listserv.fysik.dtu.dk
Subject: [ase-users] Genetic algorithm- new structures

Dear all,
Hello.

I am using the ASE genetic algorithm and quantum espresso as the calculator.
I run the code in parallel on marconi100.
1. Recently, when I run the code the genetic algorithm starts relaxing the older structures (even structures that have been already successfully relaxed) instead of creating new structures, using the same IDs. This way reaching global minima is not possible.
2. There are many cases that after 149 steps, jobs are stopped with memory error 'slurmstepd: error: Detected 1 oom-kill event(s) in step 1872160.batch cgroup'. A log file showing the large forces is attached.

I would appreciate your help.

--
Mina Taleblou
[https://my-email-signature.link/signature.gif?u=483990&e=126354219&v=b3a6b9e9699071245f95ca4fc597a90723b0ae408bcc351ec8fe6ddc71545508]



More information about the ase-users mailing list